Channel Estimation and Beamforming Techniques in Large MIMO Systems

Abstract

Wireless communication has been developed rapidly in recent years with the utilization of multiple-input multiple-output (MIMO) systems. One promising technique to support ultra-high data rate communication is massive MIMO, or large MIMO, where the number of antennas goes to tens or hundreds. Unmanned Aerial Vehicles (UAVs) have the advantages of high mobility and flexibility that can be used as aerial nodes to provide communication to ground users or base stations.

In this thesis, the optimization of pilot based channel estimation with limited pilot length is studied for a massive MIMO system. The pilot length is optimized to maximize the system spectral efficiency. Then, the optimal pilot design is discussed under the constraint of finite pilot length. The performance of uplink signal-to-interference-plus-noise ratio (SINR) and the effect of feedback error are also analyzed in time division duplex (TDD) and frequency division duplex (FDD), respectively.

Then, a large MIMO system is utilized for UAV communication. In order to provide directional beamforming, the channel estimation in UAV millimeter wave (mmWave) system is studied. A beam training and tracking method is proposed with user mobility. For beam training, a training codebook is designed based on user location distribution. For beam tracking, two tracking methods are proposed based on different types of user mobility to reduce the training overhead. The proposed beam training codebook provides larger average downlink capacity than the conventional codebook. The proposed beam tracking design is also shown to outperform the existing methods.

Lastly, a multiple UAV system in mmWave band is considered and studied. Hybrid beamforming is designed for both fully connected antenna array and partially connected subarray structures. Furthermore, a simultaneous beam tracking scheme for multiple UAV users is proposed based on subarray structure, where both UAV mobility and instability effects are taken into consideration. Simulation results show that with the proposed method, the average downlink sum capacity is improved.